On the numerical treatment and dependence of thermohaline mixing in red giants

Abstract

In recent years much interest has been shown in the process of thermohaline\nmixing in red giants. In low and intermediate mass stars this mechanism first\nactivates at the position of the bump in the luminosity function, and has been\nidentified as a likely candidate for driving the slow mixing inferred to occur\nin these stars. One particularly important consequence of this process, which\nis driven by a molecular weight inversion, is the destruction of lithium. We\nshow that the degree of lithium destruction, or in some cases production, is\nextremely sensitive to the numerical details of the stellar models. Within the\nstandard 1D diffusion approximation to thermohaline mixing, we find that\ndfferent evolution codes, with their default numerical schemes, can produce\nlithium abundances that differ from one another by many orders of magnitude.\nThis disagreement is worse for faster mixing. We perform experiments with four\nindependent stellar evolution codes, and derive conditions for the spatial and\ntemporal resolution required for a converged numerical solution. The results\nare extremely sensitive to the timesteps used. We find that predicted lithium\nabundances published in the literature until now should be treated with\ncaution.\n